How 'computer vision' could change healthcare, retail and more

There are a lot of cats on YouTube. In fact, “cat videos” have become the standard running joke about what social media is really for. So in 2012, when Google used a neural network to look at 10 million videos on YouTube and find cats, the neural network found plenty. In fact, it was 74.8% accurate when identifying cats.

The remarkable part of this experiment was that the system had not been given any information on cats’ distinguishing features — whiskers, paws, tail — that would help in identification. “It basically invented the concept of a cat,” according to the Google Fellow who led the study.

The big question for computer vision, a subset of artificial intelligence (AI) that uses deep learning, is: can a computer emulate human vision and the complex processes that take place in our brains when we see things? The current answer is: computer vision is going way beyond identifying cats.

Computer vision enables computers to see and observe so that they can derive information and make recommendations. It’s the old saying “a picture is worth a thousand words” in action. That picture is worth millions of megabytes of data for organizations looking to gain significant results and newfound efficiencies.

Instead of training machines to master the art of conversation via a chatbot, computer vision teaches machines to see things in images by breaking them down into pixels and then identifying matches. Computer vision is based on two important questions: what do I do with my eyes? What am I looking for?

Think of the need for hotels and retail outlets to ensure that they are meeting new brand standards or complying with regulations. Images from cameras could identify whether the correct signage was installed at the front desk. They could capture and review hundreds of stores to identify whether the height of a counter complies with legislation on accessibility.

Drones could take pictures of a plane looking for damage, or of electrical poles in remote areas that may start a fire. Warehouses could tell when stock is getting low without having to send people around with clipboards. Self-driving cars already use computer vision to detect objects, lanes, signs and traffic signals.

Computer vision is used in sports, manufacturing, agriculture, hospitality and facial recognition technology. Facial recognition technology is used for numerous everyday activities: unlocking phones and checking bank accounts, for example. Businesses can use facial recognition that includes emotion recognition to head off an unpleasant customer experience: for example, if someone is at a self-checkout counter and starts to have trouble, their face will typically register frustration.

A camera captures that image and alerts a customer service representative to help. It could also be used to track the metrics of smiling customers at the front desk of a hotel, so executives can see when customer satisfaction dips and identify the root cause.

Health care is a field for which computer vision is providing significant benefits: 90% of all medical data is image-based. What had to be done by people’s eyes looking at MRI images or X-rays in years past can now be analyzed by machines in a faster and more accurate way.

Rare or early-onset disease can be hard to detect for even the top medical experts. However, when machines are comparing hundreds of thousands of images, anomalies that are hard for the human eye to detect can be brought to the attention of a doctor. Such insights can bring dramatic positive outcomes to patients and the health care organizations that support them. Disease identified early can save time, money, and lives.

Computer vision is being used today in content moderation on social media platforms, helping to detect and remove toxic or harmful content. In 2022, social networking sites are estimated to reach 3.96 billion users, a number expected to grow as mobile device usage and mobile social networks continue to expand in previously underserved markets.

Every day, more than 300 million photos are uploaded to Facebook, the largest social media platform, and every minute there are 510,000 comments posted and 293,000 status updates. The majority of content is benign (and may even include cute cats) but enough of it is considered harmful that Facebook alone now has 15,000 content moderators. A recent report estimates that the company’s AI system and human beings flag more than three million items daily.

Computer vision can be used for video, image and text moderation, significantly reducing the burden on human beings and more quickly and efficiently identifying the most egregious content. Computer vision can help ensure accuracy, as it doesn't get tired eyes or general fatigue, which means it's less likely to make mistakes and more likely to find the critical parts of an image — such as a flaw in a product on the production line — with precision and speed. Machines don’t get bored and are programmed for long hours with non-stop performance.

The ethics of computer vision

As with all AI, computer vision doesn’t have moral values and can’t make judgments on what is ethical or legal. It needs strict monitoring by humans to make sure it’s safe, reliable, and not compromising privacy or being used for nefarious purposes. Organizations should use the technology in an ethical manner, using frameworks such as Trustworthy AI^TM, preparing for government regulations related to AI, and mitigating their risks.

Despite the risks, computer vision is here to stay. Automating “sight” means employees can focus on more strategic activities that can drive positive customer experience, optimize jobs, and even save lives. By combining human and computer vision, organizations could achieve goals that they would never achieve otherwise.